FAQ

Frequently asked questions about Chat Analyzer's text preprocessing, supported formats, privacy principles, and how to interpret AI analysis results.

1. What exactly does Chat Analyzer analyze—and what does it NOT measure?

After stripping date dividers and leading timestamps and merging consecutive messages from the same speaker, it generates a nickname, a brief comment for each participant, and a one-line summary based on observable phrasing and role balance. It does NOT measure average reply times, read-receipt status, attraction percentages, psychological attachment types, or MBTI personality accuracy.

2. Are my uploaded chat logs stored on a server or database?

No. Chat Analyzer does not run any database to store your original chat logs or AI analysis results. However, when you click Analyze, the preprocessed text is transmitted through Cloudflare servers to the Google Gemini API for processing, so you should remove or mask real names, phone numbers, addresses, and account numbers beforehand.

3. Is there an automatic anonymization feature for names and phone numbers?

No automatic masking is provided because regex alone cannot guarantee 100% removal of context-embedded proper nouns or private details. Before uploading, please use Find & Replace (Ctrl+H) in a text editor to replace names with “A” or “B” and delete any sensitive lines.

4. Do Mild Mode and Spicy Mode use different analysis algorithms?

Both modes use the exact same text preprocessing and the same Gemini model, but differ in the system prompt’s interpretive lens and tone. Mild Mode highlights consideration, cooperation, and concrete proposals in a warm voice, whereas Spicy Mode points out habits like one-sided questioning or short replies with wit and directness. A difference in tone does not mean a difference in accuracy.

5. If I upload years of chat history, does it analyze the entire file from start to finish?

No. After merging consecutive bubbles from the same speaker, it slices up to the most recent 3,000 turns (within 20,000 characters) for analysis. If you want to analyze a specific conversation or conflict from the past, copy just that excerpt into a separate .txt file and upload it.

6. Can it analyze group chats with 3 or more participants?

Yes. The parser automatically identifies all speakers in the dialogue. For group chats with 3 or more people, result cards are split into groups of up to 3 participants per card, and each person’s comment is kept within 80 characters (or concise two-sentence limits) for readability.

7. How do I export chat logs on Android, iPhone (iOS), and PC?

On Android KakaoTalk, open Chat Settings → Export Chat → Send Text Only, then save to your device or share directly to the installed web app. On iPhone (iOS), choose Export Chat → Text Only → Save to Files, then upload the .txt file from your browser. On PC, use the chat menu → Export Chat to save a .txt file directly.

8. Why do I get an “Unsupported file encoding” or “Could not recognize conversation format” error?

Encoding errors happen when a file uses legacy encodings (such as EUC-KR/CP949) instead of UTF-8, leaving broken characters (U+FFFD). Open the file in Notepad, choose Save As, and select UTF-8 encoding. Format errors occur when speaker delimiters (e.g., “A: Hello” or “[A] Hello”) are missing or the text has fewer than 10 characters.

9. Does it work with LINE, Instagram DM, or plain text logs besides KakaoTalk?

Yes. As long as the file is a UTF-8 encoded .txt file where each line separates speaker and message with a colon (“A: message”), brackets (“[A] message”), or a tab, it can be parsed and analyzed regardless of the messenger.

10. Why did the AI mention a relationship label or personality trait not stated in the chat?

Large Language Models (LLMs) sometimes fill in narrative gaps from brief context (hallucination)—for example, assuming two people planning a weekend walk are a couple. Filter out any background labels or personality claims that lack direct textual evidence and focus only on the observable actions.

11. If one person sends five short bubbles in a row, does that inflate their share?

Counting raw bubbles would overstate rapid-fire texters, so Chat Analyzer merges consecutive messages from the same speaker into a single Turn during preprocessing. Sending five short bubbles or one combined paragraph counts as the exact same single turn.

12. Why are multiple PNG images downloaded when saving result cards?

To preserve a clean square aspect ratio optimized for mobile screens and social sharing, each result card displays up to 3 participants. In group chats with 4 or more people, cards are split sequentially and downloaded as separate PNG files.

13. Is it free to use? Why do I see a rate-limit notice?

Yes, all features are free with no sign-up or payment required. To prevent server overload from excessive automated or repeated API calls, requests are rate-limited per IP (up to 3 per minute and 5 per day). Please wait a moment and try again if you reach the limit.

14. Is it okay to share the analysis results directly with the other person?

Sharing results as a lighthearted icebreaker can be fun, but we do not recommend using Spicy Mode nicknames or blunt comments as “proof” during an argument. When sharing result images publicly or in group chats, always get the participants’ consent and verify that no real names or private context are exposed.

Still have questions?

Check our How to Use guide for step-by-step export instructions, or reach out via Contact for support.